zain khan

zain khanSep 18, 2026

"This has already been done. What's the point?" Eman Asif & I heard some version of that a lot while working on our final year project. Making dark photos clear isn't a new problem. People have been working on it for years. For a while it did get to us . Everyone around us seemed to be building something new, and we felt like, we were re-treading old ground. We built it anyway. And I'm glad we did. BioNAFNet takes a different route, instead of copying what other models do, we asked how the human eye handles darkness and built that into the network. Your eyes do several clever things in the dark. Rod cells take over from cone cells. Neighbouring cells push against each other to sharpen edges. Your pupils widen to gather more light. Your brain handles brightness and colour on separate channels. And you notice small changes in dark areas far more than in bright ones... We turned seven of those mechanisms into working components of a neural network. That's the part nobody had done before. And thats what came out of it, our model recover images more accurately than well knows published methods like Retinex Former andMIR-Net etc on standard benchmarks What we actually learned: "someone already solved this" and "someone already solved this the way you're about to" are two different sentences. The second is a real problem. The first is just the field existing. To anyone second guessing their FYP because the topic isn't brand new , the contribution is in the how, not the headline. Worth saying, this was a 50/50 project with Eman Asif. Months of training runs, benchmarking and debugging, split between us. Solid work partner.

zain khanSep 7, 2026

A photo from a conference room in Newcastle made my week. Our paper, "From Sensors to Prediction: Comparing Learning Models for Sustainable Cooling Systems," was presented at the 19th UK Heat Transfer Conference at Newcastle University . Seeing my name on that slide, coming from a lab in Lahore, is a moment I won't forget. The paper looks at how different learning models perform when predicting sustainable cooling system behaviour from sensor data. It came out of the UKRI-funded S2Cool project, a joint effort between LUMS and Newcastle University. Special thanks to Mahar Zeeshan who led this work and carried it to Newcastle, and to Asim Karim, whose supervision shaped how we approach research. I learned a great deal from both of them on this project. #Research #MachineLearning #SustainableCooling #UKRI

zain khanSep 21, 2026

11:47 PM. Lights off. Laptop closed. I was done for the day. Then, just before sleeping, I put on the Interstellar main theme. . By the time it built up, I wasn't in my room anymore. I was somewhere between the stars, thinking about time, about how small we are, and how far curiosity can take us. I opened my eyes and grabbed my phone. Not to scroll, but to write. One idea. Then another. Then five more. It was 1:30 AM when I finally put my phone down..... I hadn't solved anything. I hadn't built anything..... But I had a notes app full of ideas and a feeling I hadn't had in a while... Sometimes inspiration doesn't come from a lecture, a book, or a meeting.... Sometimes it comes from soundtracks and a sleepless night....